Kong API Gateway Interview Questions and Answers (2027)

80 Kong API Gateway interview questions with answers: services, routes, plugins, decK, DB-less and hybrid mode, Kong Ingress Controller, Konnect.

Oct 1, 2025 - 11:39
Nov 7, 2026 - 00:05
 0  482
Kong API Gateway Interview Questions and Answers (2027)

Kong API Gateway interview questions check whether you can put Kong Gateway in front of real APIs: model Gateway Services, Routes, Upstreams and Consumers, attach plugins for authentication and rate limiting, manage configuration as code with decK, choose between traditional, DB-less and hybrid mode, and run Kong on Kubernetes with the Kong Ingress Controller or through Kong Konnect. This page has 80 questions with direct answers, grouped by experience level.

Product details were checked against the Kong documentation at developer.konghq.com in November 2026. At that point Kong Gateway 3.16 was the latest release, and 3.10 and 3.14 were the long-term support (LTS) versions.

Key takeaways

  • Freshers are asked about the core entities (Gateway Service, Route, Upstream, Target, Consumer, Plugin), the Admin API and the default ports.
  • Candidates with 2 to 5 years of experience should explain DB-less and hybrid mode, write a declarative configuration file, run deck gateway diff and deck gateway sync in a pipeline, and know how plugin precedence works.
  • Senior candidates are tested on hybrid mode certificates, CP and DP version rules, rate limiting accuracy across nodes, upgrades between LTS versions, custom plugins and Admin API security.
  • On Kubernetes, expect questions on KongPlugin, the konghq.com/plugins annotation, Gateway API HTTPRoute support and the deprecation of KongIngress.
  • Kong Konnect hosts the control plane for you and adds Dev Portal, Observability, Catalog, Event Gateway and AI Gateway features. Know what it changes compared with a self-managed control plane.
  • Scenario questions (404 "no Route matched", 502 and 503 upstream errors, inconsistent rate limits, a DP that stops receiving config) carry the most weight for experienced roles.

How to use this guide

Each answer opens with the reply you could give in an interview, then adds the detail a follow-up question would probe. If you are preparing for wider DevOps rounds too, the 200+ DevOps interview questions guide covers Linux, Git, CI/CD and Kubernetes around the gateway topics here.

Your experienceRead firstThen practise
Fresher or career switcherQuestions 1 to 25Run Kong in Docker, create a Service, a Route and a key-auth Consumer with curl
2 to 5 yearsQuestions 26 to 58Manage the same setup with a decK state file and a pipeline that runs diff before sync
Senior, lead or architectQuestions 59 to 70Hybrid mode with PKI certificates, a Redis-backed rate limit and an upgrade plan
Everyone with work experienceQuestions 71 to 80 (scenarios)Explain each scenario aloud: symptom, check, fix, prevention

Kong API Gateway interview questions for freshers

1. What is Kong Gateway?

Kong Gateway is an API gateway that sits between clients and your backend services. It receives every request, matches it to a Route, runs plugins such as authentication, rate limiting and logging, and proxies the request to the right upstream service. It is built on OpenResty (NGINX plus LuaJIT), and most plugins are written in Lua. You can run it self-managed or with a control plane hosted in Kong Konnect.

2. What problems does an API gateway solve in a microservices setup?

It gives clients one entry point and moves shared concerns out of each service. Without a gateway, every team writes its own authentication, rate limiting, CORS handling and request logging. With Kong, those concerns become plugins configured once and applied per Route, Service or Consumer, while the services keep only business logic. Our post on the role of API gateways in securing microservices covers the security side in more depth.

3. What are the core Kong Gateway entities?

EntityWhat it represents
Gateway ServiceAn upstream application that holds your business logic
RouteRules that match incoming requests (paths, hosts, methods, headers) and send them to a Service
UpstreamA virtual hostname that load balances across a set of Targets
TargetOne backend instance, an IP or hostname plus a port
ConsumerAn external client of your APIs, such as a partner or a mobile app
Consumer GroupA set of Consumers that share settings like rate limits
PluginA module that adds behaviour to request and response processing
Certificate, SNI, CA CertificateTLS material for serving and verifying certificates
VaultA secret store that other entities reference instead of holding plaintext secrets
WorkspaceAn isolation boundary for entities (Enterprise and Konnect)

4. What is a Gateway Service?

A Gateway Service is Kong's record of one backend API. It holds the protocol, host, port and path Kong uses to reach that backend, plus settings such as retries and connect, read and write timeouts. A Service can point straight at a hostname or at an Upstream name when you want Kong to load balance across several Targets.

5. What is a Route?

A Route decides which requests go to which Service. It matches on attributes such as paths, hosts, methods, headers and, for TLS traffic, snis. One Service usually has several Routes, for example a public path and an internal one with different plugins.

6. What do strip_path and preserve_host do on a Route?

strip_path removes the matched path prefix before Kong forwards the request. With paths: [/orders] and strip_path: true, a call to /orders/42 reaches the backend as /42. preserve_host forwards the client's original Host header instead of the Service's hostname, which matters when the backend routes by host name.

7. What are Upstreams and Targets?

An Upstream is a named, virtual backend that Kong load balances across. Each Target inside it is an address:port with a weight. You point a Service's host at the Upstream name, and Kong spreads traffic across healthy Targets. Changing Target weights is a simple way to shift traffic between versions.

8. What is a Consumer in Kong?

A Consumer represents whoever calls your API: an application, a partner or a user. Authentication plugins attach credentials (an API key, a JWT secret, basic auth details) to a Consumer, and once Kong identifies the Consumer it can apply per-Consumer plugins such as a dedicated rate limit or an ACL group.

9. What is a Kong plugin?

A plugin is a module that runs during request or response processing. Kong ships plugins for authentication (key-auth, jwt, basic-auth, oauth2), traffic control (rate-limiting, request-size-limiting, proxy-cache), security (ip-restriction, cors, bot-detection), transformation (request-transformer, response-transformer) and observability (prometheus, opentelemetry, http-log). You enable a plugin globally or scope it to Services, Routes, Consumers or Consumer Groups.

10. What is the Admin API, and how do you create a Service and Route with it?

The Admin API is Kong's REST interface for configuration. It listens on port 8001 (HTTP) and 8444 (HTTPS) by default. In traditional mode, every change you make through it is stored in the database and picked up by all nodes.

curl -i -X POST http://localhost:8001/services \
  --data name=orders \
  --data url=http://orders.internal:8080

curl -i -X POST http://localhost:8001/services/orders/routes \
  --data name=orders-route \
  --data 'paths[]=/orders'

curl -i -X POST http://localhost:8001/services/orders/plugins \
  --data name=rate-limiting \
  --data config.minute=60 \
  --data config.policy=local

11. Which ports does Kong use by default?

PortPurpose
8000 / 8443Proxy traffic from clients (HTTP / HTTPS)
8001 / 8444Admin API (HTTP / HTTPS)
8002 / 8445Kong Manager GUI (HTTP / HTTPS)
8005Hybrid mode: control plane listens for data plane connections
8006Hybrid mode: telemetry from data planes
8007Status listener for monitoring (3.6 and later)

Kong's docs advise blocking public access to the Admin API ports in production.

12. What is Kong Manager?

Kong Manager is the web GUI for Kong Gateway, served on port 8002 by default. It lets you view and edit Services, Routes, Consumers and plugins. In Enterprise it adds Workspaces and RBAC-controlled access. In DB-less mode it is read-only, because the Admin API behind it cannot write.

13. Is Konga part of Kong?

No. Konga is a community-built admin UI that was popular with older Kong versions. It is not a Kong product and Kong does not support it. For current versions, use Kong Manager, Kong Konnect or decK.

14. What is the difference between Kong Gateway open source and Kong Gateway Enterprise?

Both share the same core proxy, entities and many bundled plugins. Enterprise adds features aimed at larger organisations, such as Workspaces, RBAC, enterprise-only plugins (for example OpenID Connect and the advanced variants of rate limiting, proxy caching and transformers), secrets manager integrations and vendor support. Before choosing the open source build for production, check Kong's current packaging and support terms, since they have changed over time.

15. How do you protect an API with key authentication?

Enable the key-auth plugin on the Route or Service, create a Consumer, and give that Consumer a key. Clients then send the key in a header (default name apikey) or query parameter. Requests without a valid key get 401.

curl -X POST http://localhost:8001/routes/orders-route/plugins --data name=key-auth
curl -X POST http://localhost:8001/consumers --data username=mobile-app
curl -X POST http://localhost:8001/consumers/mobile-app/key-auth
curl -i http://localhost:8000/orders -H "apikey: <key-from-previous-step>"

16. Which authentication plugins does Kong offer?

Common ones are key-auth, basic-auth, jwt, hmac-auth, ldap-auth and oauth2. Enterprise adds plugins such as openid-connect and mtls-auth. For most new designs with an external identity provider, OpenID Connect is the usual choice, because the IdP issues and rotates tokens instead of Kong.

17. What does the ACL plugin do?

The acl plugin allows or denies access based on groups assigned to a Consumer. It needs an authentication plugin on the same Route or Service so Kong knows who the Consumer is. A typical setup puts partner Consumers in a partners group and allows only that group on a partner Route.

18. How does the rate-limiting plugin respond when a client exceeds its limit?

Kong returns HTTP 429 with the body {"message":"API rate limit exceeded"} and a Retry-After header. On normal responses it adds headers such as RateLimit-Limit, RateLimit-Remaining, RateLimit-Reset and per-window headers like X-RateLimit-Remaining-Minute. You can hide these with config.hide_client_headers.

19. What is the CORS plugin for?

The cors plugin answers browser preflight requests and adds the Access-Control-Allow-* headers, so a web front end on one origin can call an API on another. You list allowed origins, methods and headers in its config instead of coding CORS into every backend.

20. When would you use request-transformer and response-transformer?

Use them to add, remove, rename or replace headers, query parameters and simple body fields without changing backend code. Examples: add an internal header that tells the backend which gateway route was used, strip a server version header from responses, or rename a query parameter during an API migration.

21. What do ip-restriction and bot-detection do?

ip-restriction allows or denies requests by client IP or CIDR range, which suits admin or partner endpoints. bot-detection blocks requests whose User-Agent matches known bot patterns or your own deny list. Neither replaces a full WAF or bot management product; they are first-line filters.

22. What is the request-termination plugin used for?

It makes Kong answer a request itself with a fixed status code and message instead of proxying it. Teams use it to take a Route offline for maintenance, return 410 for a retired endpoint, or stub an endpoint during testing.

23. Which logging plugins does Kong provide?

Kong can send per-request logs with http-log, tcp-log, udp-log, file-log, syslog, loggly and others, plus metrics through statsd, datadog and prometheus. These run in the log phase, after the response is sent, so they add little to client latency.

24. How do you monitor Kong with Prometheus?

Enable the prometheus plugin, usually globally, and scrape the metrics endpoint Kong exposes (the Status API is a common choice so the Admin API stays private). You get request counts, latency histograms and upstream health metrics, which you can graph in Grafana.

25. How do you configure a Kong node?

Through kong.conf, or through environment variables with the KONG_ prefix, which is how most container deployments do it. For example, KONG_DATABASE=off sets database = off. Run kong check <conf> to validate a configuration file before starting the node.

Kong interview questions for 2 to 5 years of experience

26. What deployment topologies does Kong Gateway support?

TopologyHow it worksWhere it runs
Traditional (database) modeEvery node connects to the database and acts as both control plane and data planeSelf-managed only
DB-less modeEntities live in memory, loaded from a declarative YAML or JSON fileSelf-managed only
Hybrid modeControl plane nodes hold config (with a database); data plane nodes run without one and receive config from the CPSelf-managed or Konnect-hosted CP

Traditional mode is the only topology that supports database-dependent plugins such as OAuth2 or rate limiting with the cluster policy.

27. What is DB-less mode, and what are its limits?

In DB-less mode you set database = off and give Kong a declarative file, either at startup with declarative_config or at runtime through the /config Admin API endpoint. The limits Kong documents:

  • The Admin API is read-only; write requests to entity endpoints return 405.
  • The whole configuration must fit in the cache (mem_cache_size, default 128m).
  • Nodes are independent, so you load config into each node.
  • Kong Manager cannot create or change entities, and deck gateway sync cannot be used against the node.
  • Plugins that store data in the database are not fully compatible.

28. What does a declarative configuration file look like?

_format_version: "3.0"
services:
  - name: orders
    url: http://orders.internal:8080
    tags: [team-orders]
    routes:
      - name: orders-route
        paths:
          - /orders
        strip_path: true
        plugins:
          - name: key-auth
    plugins:
      - name: rate-limiting
        config:
          minute: 60
          policy: local
consumers:
  - username: mobile-app
    tags: [team-orders]

The same format is used for DB-less files and decK state files. Keep credentials out of the file and inject them at deploy time.

29. Which kong config commands should you know?

  • kong config init writes an example kong.yml.
  • kong config parse kong.yml checks a declarative file's syntax without loading it.
  • kong config db_import kong.yml loads a declarative file into the database.
  • kong config db_export exports the database to a declarative file.

30. What is decK?

decK is Kong's command-line tool for managing configuration declaratively. It talks to the Admin API (or to a Konnect control plane) and compares a state file with what is running. The core commands are deck gateway ping, deck gateway dump, deck gateway diff, deck gateway sync and deck gateway validate. File commands such as deck file openapi2kong and deck file merge transform state files offline.

31. How would you use decK in a CI/CD pipeline?

Keep the state file in Git, validate and diff on every pull request, and sync only after merge. Reviewers then see exactly which entities will change.

# pull request job
deck gateway ping
deck gateway validate kong.yaml
deck gateway diff kong.yaml

# main branch job, after approval
deck gateway sync kong.yaml

For a Konnect control plane, add --konnect-control-plane-name and a token (decK v1.40.0 or later). The same pattern fits any of the pipeline styles in our CI/CD pipeline examples.

32. How do select tags help when several teams share one Kong cluster?

Tag each team's entities and run decK with --select-tag. decK then only reads, diffs and syncs entities carrying that tag, so one team's sync does not delete another team's Routes. Without tags, sync treats the state file as the full truth and removes anything missing from it.

33. What is APIOps with deck file openapi2kong?

APIOps means generating gateway config from the API specification instead of writing it by hand. deck file openapi2kong -s oas.yaml -o kong.yaml turns an OpenAPI document into Kong Services and Routes. A pipeline can then merge in team plugins with deck file merge, diff and sync, so the gateway always matches the published spec.

34. What is hybrid mode?

Hybrid mode splits Kong into control plane (CP) nodes and data plane (DP) nodes. You set role = control_plane or role = data_plane. Only CPs need a database. You configure the CP through the Admin API, Kong Manager or decK, and it pushes configuration to DPs, which run DB-less and proxy traffic. DPs can sit in many regions or clouds without database access.

role = data_plane
database = off
cluster_control_plane = cp.example.internal:8005
cluster_telemetry_endpoint = cp.example.internal:8006
cluster_mtls = pki
cluster_cert = /etc/kong/cluster.crt
cluster_cert_key = /etc/kong/cluster.key
cluster_ca_cert = /etc/kong/ca.crt

35. What is the difference between shared and pki in cluster_mtls?

With shared, every CP and DP uses the same certificate and key pair. It is quick to set up, but one leaked key exposes the whole cluster. With pki, each DP has its own certificate issued by a common CA given in cluster_ca_cert, so you can revoke one node. pki_check_cn adds a common name check against cluster_allowed_common_names.

36. What version rules apply between control plane and data planes?

On self-managed hybrid mode, the CP only accepts DPs with the same major version. It rejects DPs with a newer minor version but accepts newer patch versions. Plugin versions follow a similar rule. If a check fails, the CP stops pushing new config to that DP and logs a warning. In practice: upgrade the CP first, then the DPs.

37. What happens to data planes if the control plane goes down?

DPs keep proxying with their last known configuration and keep trying to reconnect. They can even restart while disconnected, because they cache config locally (by default in an LMDB file, dbless.lmdb, under the Kong prefix, which you can encrypt). You cannot change configuration until the CP returns.

38. What are the rate-limiting policies, and when do you use each?

PolicyWhere counters liveUse when
localIn memory on each nodeOne node, or approximate limits are fine; fastest option
clusterKong's databaseTraditional mode only; not supported in hybrid mode or Konnect
redisA shared Redis serverSeveral nodes need accurate, shared counters

The Rate Limiting Advanced plugin adds multiple limits with sliding or fixed windows and Redis Sentinel or Redis Cluster support.

39. If the same plugin is configured at several scopes, which one runs?

Only one instance runs per request, and the most specific scope wins. Kong's order, from highest to lowest precedence, starts with Consumer plus Route plus Service, then Consumer Group plus Service plus Route, and continues down to Route, then Service, then global. So a rate limit set on a specific Consumer and Route overrides a global rate limit for that Consumer.

40. Which phases can a plugin run in?

For HTTP traffic the main phases are certificate (TLS handshake), rewrite (global plugins only, before Kong identifies the Service), access (before proxying, where auth and rate limits run), response or header_filter and body_filter (response processing), and log (after the last byte is sent). Stream traffic uses preread and log. Knowing the phase explains why a plugin cannot, for example, rewrite a request after it reached the upstream.

41. How does Kong decide the order of different plugins?

Each plugin has a static priority number, and higher priority plugins run first within a phase. That is why authentication runs before rate limiting by default. Kong Gateway Enterprise also supports dynamic plugin ordering, which lets you tell a plugin instance to run before or after another one in the access phase, for example running rate limiting before authentication to shed load from unauthenticated floods.

42. Which load-balancing options does an Upstream offer?

Round-robin is the default. You can also choose consistent hashing (by consumer, IP, header, cookie or path), which keeps a client on the same Target, and least-connections, which favours the Target with the fewest active connections. Target weights work with all of them.

43. How do active and passive health checks differ?

Active health checks probe each Target on a schedule (for example GET /health) and mark it healthy or unhealthy by status code and timeouts. Passive health checks, also called circuit breakers, watch real proxied traffic and mark a Target unhealthy after a number of failures. Passive checks cannot mark a Target healthy again by themselves, so teams often combine both.

44. How do you run a canary release with Kong?

The simplest way is two Targets in one Upstream, for example v1 at weight 90 and v2 at weight 10, then shift the weights while you watch error rates. You can also create a separate Route for v2 matched by a header so testers reach it on purpose. Kong Gateway Enterprise also has a dedicated canary plugin.

45. What are Consumer Groups for?

They let you apply one plugin configuration to many Consumers at once. A common use is tiered plans: put free-tier Consumers in one group with a low rate limit and paid Consumers in another with a higher limit, instead of configuring each Consumer.

46. How does the JWT plugin work?

The jwt plugin checks that the request carries a JWT signed by a credential registered on a Consumer. Kong reads a claim (by default iss) to find the matching credential, verifies the signature with its secret or public key, and can check exp and nbf. It verifies tokens; it does not issue them.

47. OAuth2 plugin or OpenID Connect plugin?

The bundled oauth2 plugin makes Kong act as the authorization server and stores tokens in Kong's database, so it needs traditional mode. The Enterprise openid-connect plugin works with an external identity provider such as Keycloak, Okta or Entra ID, validates its tokens and supports several flows. With an existing IdP, OpenID Connect is usually the better fit.

48. How do you secure traffic between Kong and the upstream?

Use https as the Service protocol, turn on tls_verify and attach the trusted CA through ca_certificates so Kong checks the upstream certificate. For mutual TLS, set a client_certificate on the Service so Kong presents its own certificate to the backend.

49. How should secrets be referenced in Kong configuration?

Use the Vault entity and secret references instead of plaintext values. A reference looks like {vault://env/redis-password}, which reads an environment variable; other backends such as HashiCorp Vault or cloud secret managers use their own prefixes. Kong resolves the value at runtime, so the secret never appears in your decK file. Our Vault interview questions cover the secret store side.

50. What are Workspaces and RBAC?

Both are Enterprise features. Workspaces split one Kong cluster into isolated groups of entities, often one per team. RBAC assigns roles and permissions to Admin API and Kong Manager users, so a team can manage its own Routes and plugins without touching other workspaces.

Kong on Kubernetes

51. What is the Kong Ingress Controller?

The Kong Ingress Controller (KIC) runs in your cluster, watches Kubernetes resources such as Ingress, HTTPRoute, Services and Kong custom resources, and converts them into Kong Gateway configuration. Developers then manage gateway config with the same manifests and Helm charts they use for their apps.

52. How does KIC push configuration to Kong?

In the usual setup KIC builds a full configuration and sends it to each Kong Gateway pod through the /config endpoint, so the gateways run DB-less. If a new configuration is rejected, Kong keeps the last known good config. A DB-backed topology also exists, where KIC configures a control plane that uses PostgreSQL.

53. How do you attach a plugin to a route on Kubernetes?

Create a KongPlugin resource and reference it with the konghq.com/plugins annotation on an Ingress, Service, HTTPRoute, KongConsumer or KongConsumerGroup.

apiVersion: configuration.konghq.com/v1
kind: KongPlugin
metadata:
  name: rate-limit-60
plugin: rate-limiting
config:
  minute: 60
  policy: local
---
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: orders
  annotations:
    konghq.com/plugins: rate-limit-60
spec:
  parentRefs:
    - name: kong
  rules:
    - matches:
        - path:
            type: PathPrefix
            value: /orders
      backendRefs:
        - name: orders
          port: 8080

54. What is the difference between KongPlugin and KongClusterPlugin?

KongPlugin is namespaced and can only be referenced from resources in the same namespace. KongClusterPlugin is cluster-scoped, so a platform team can define one plugin (for example a global logging plugin) and let resources in any namespace use it.

55. Does KIC support the Kubernetes Gateway API?

Yes. KIC treats Gateway API resources as native and converts them straight into Kong configuration. Kong's docs state that KIC 3.5 and later conform to Gateway API v1.3, and that it handles HTTPRoute, TCPRoute and UDPRoute. For new clusters, Gateway API is usually preferred over the older Ingress resource.

56. Is KongIngress still used?

It is deprecated. Kong replaced it with annotations and the KongUpstreamPolicy resource for upstream settings such as load-balancing algorithm and health checks. TCPIngress and UDPIngress are also deprecated in favour of Gateway API routes. If an interviewer shows you a manifest with KongIngress, say how you would migrate it.

57. What is the Kong Operator?

The Kong Operator (also called Kong Gateway Operator) is an optional component that provisions and manages Kong Gateway deployments declaratively. With it, you get GatewayClass and Gateway driven provisioning, GRPCRoute support and a KongLicense resource for Enterprise licences.

58. How is Kong Mesh different from Kong Gateway?

Kong Gateway handles north-south traffic: requests entering your platform from outside. Kong Mesh is a service mesh, built on the open source Kuma project, for east-west traffic between services, with sidecar proxies providing mTLS, retries and traffic policies. Many platforms run both. Compare it with Istio using our Istio interview questions.

Senior Kong interview questions

59. What is Kong Konnect?

Konnect is Kong's SaaS management platform. It hosts the control plane and offers API Gateway management, Dev Portal, Konnect Observability, Catalog, Service Mesh management, Event Gateway (a Kafka protocol proxy), AI Gateway, Dedicated Cloud Gateways, a Serverless gateway option and Metering & Billing. You run data planes yourself or let Kong run them.

60. When would you pick Dedicated Cloud Gateways over self-managed data planes?

Dedicated Cloud Gateways are data plane nodes that Kong runs for you in AWS, Azure or GCP. Pick them when the team does not want to patch and scale gateway nodes. Keep self-managed DPs when traffic must stay in your own network, you need custom plugins or kernel tuning Kong does not offer as a managed option, or regulations require you to control the runtime.

61. How does Kong's version support work, and which versions are current?

Kong releases minor versions regularly and marks some as LTS with a longer support window. In November 2026, Kong's support policy listed 3.16 as the latest minor (released 15 September 2026, supported to 15 September 2027), 3.14 as LTS (supported to 7 April 2029) and 3.10 as LTS (supported to 31 March 2028). Production teams usually standardise on an LTS line.

62. How would you upgrade a traditional-mode Kong cluster?

Read the changelog and breaking changes, back up the database, and test the new version in staging with your real decK file. Kong's migration commands are kong migrations up, which runs new migrations while old nodes keep serving, and kong migrations finish, which completes them after old nodes are gone. A blue/green approach (new nodes on the new version, shift traffic, then finish migrations) keeps downtime near zero. In hybrid mode, upgrade CPs before DPs.

63. How do you write a custom plugin?

A Lua plugin has at least a handler.lua, which implements phase functions such as access and sets PRIORITY and VERSION, and a schema.lua, which defines and validates its configuration. Inside the handler you use the Plugin Development Kit (PDK), for example kong.request.get_header() and kong.response.exit(). Kong also has a Go PDK. Install the plugin on every node and add it to the plugins setting in kong.conf.

64. What would you check to tune Kong performance?

  • Plugin cost: plugins that call external services (introspection, external auth) add network round trips. Cache results where the plugin supports it.
  • Rate limiting: local is fastest; redis adds a Redis call per request, so keep Redis close to the nodes.
  • Memory: in DB-less and hybrid DPs, mem_cache_size must fit the whole configuration.
  • Workers and connections: NGINX worker processes and upstream keepalive settings.
  • Logging: send logs asynchronously rather than writing large bodies.

65. How would you design observability for Kong?

Use the prometheus plugin for metrics, the opentelemetry plugin for traces and a log plugin (for example http-log) into your log platform. Add a correlation ID with the correlation-id plugin so a request can be followed from gateway to backend. If you use Konnect, Konnect Observability adds hosted analytics. Our observability tools guide compares the back ends.

66. How do you secure the Admin API?

Never expose it publicly. Bind admin_listen to a private interface or localhost, put it behind a firewall or VPN, and use HTTPS on 8444. In Enterprise, turn on RBAC so every call needs a token tied to a role. Better still, make Git and decK the only way configuration changes, so humans rarely call the Admin API directly. In hybrid mode, DPs do not serve the Admin API at all.

67. How would you design Kong for a multi-region deployment?

Hybrid mode fits well: one control plane (self-managed with a highly available database, or hosted in Konnect) and data planes in each region behind regional load balancers and global DNS or anycast. Use pki cluster certificates per region, a regional Redis for rate limiting, and decK in a pipeline as the single source of configuration. Plan for the CP being unreachable: DPs keep serving with cached config, but changes wait.

68. What is Kong AI Gateway?

It is a set of AI features built on Kong Gateway for traffic to large language models, MCP tool servers and agents. It includes policies such as AI Proxy, AI Prompt Guard, AI Semantic Cache, AI Rate Limiting Advanced (spend limits based on token cost), AI Sanitizer for PII redaction, and guardrail integrations with AWS, Azure and Google services. It also lets you expose existing APIs as MCP tools. Check the docs for your version, as this area changes quickly.

69. How does Kong compare with other gateways?

OptionStrengthsTrade-offs
Kong GatewayRuns anywhere (VMs, Kubernetes, any cloud), large plugin set, decK and Konnect for central managementYou operate it unless you use Konnect-managed data planes
AWS API GatewayFully managed, tight IAM and Lambda integrationTied to AWS; less control over runtime and plugins
Plain NGINX or an ingress controllerSimple and familiar for basic routingAuth, rate limiting and consumer management take custom work
Service mesh (Istio, Kong Mesh)mTLS and policy between servicesSolves east-west traffic; usually paired with a gateway at the edge

70. How do you test Kong configuration before production?

Run deck gateway validate and deck gateway diff against a staging cluster in the pull request, then run API tests (for example with a collection runner or contract tests) through the staging gateway. Test auth failures and rate limit responses on purpose, not only the happy path. Promote the same state file to production so staging and production cannot drift.

Scenario-based Kong interview questions

71. Clients get 404 with "no Route matched with those values". What do you check?

That message means no Route matched the request's host, path, method or headers. Compare the request with the Route definitions: a wrong hosts value, a missing leading slash, a method not listed, or a regex path missing the ~ prefix. Check you sent traffic to the proxy port (8000 or 8443), not the Admin API. In DB-less or KIC setups, confirm the new config actually loaded.

72. Kong returns 502 Bad Gateway for one service. How do you debug it?

A 502 means Kong reached for the upstream and got an invalid response or a connection failure. Check Kong's error.log (by default under /usr/local/kong/logs/) for the exact error, then test the upstream directly from a Kong node with curl. Typical causes are a wrong port, the backend speaking HTTPS while the Service says HTTP (or the reverse), TLS verification failures, or the backend closing keepalive connections early.

73. Kong returns 503 with "failure to get a peer from the ring-balancer". What happened?

All Targets in the Upstream are marked unhealthy, or the Upstream has no Targets. Check health status through the Admin API (GET /upstreams/{name}/health), fix the health check path or thresholds if they are wrong, and remember that passive checks need an active check or manual action to mark Targets healthy again.

74. A 60 requests per minute limit lets through far more than 60 across the cluster. Why?

The plugin is using the local policy, so each node counts separately. With four nodes behind a round-robin load balancer, a client can get close to 240 requests per minute. Switch to the redis policy with a shared Redis, or accept approximate limits and divide the limit by node count. The cluster policy is not an option in hybrid mode or Konnect.

75. A rate limit set on a Route is not applied to one partner. Why?

A more specific plugin instance exists. If the partner's Consumer has its own rate-limiting instance, Consumer-scoped configuration takes precedence over the Route-scoped one, and only one instance runs. List plugins for that Consumer and Route through the Admin API or decK dump and remove or adjust the overriding one.

76. A data plane stopped receiving configuration updates. Where do you look?

Check the DP and CP logs. Common causes: an expired or mismatched cluster certificate, a DP running a newer minor version than the CP (the CP stops pushing config and logs a warning), a plugin version mismatch, or a firewall blocking port 8005. Meanwhile the DP serves cached config, so traffic flows but changes do not arrive.

77. One team's decK sync deleted another team's Routes. How do you prevent a repeat?

Their state file did not contain the other team's entities, and sync removes anything not in the file. Restore from the last good dump or Git history. Then tag every team's entities, require --select-tag in each team's pipeline, or move teams into separate Workspaces (Enterprise) or Konnect control planes, and make deck gateway diff output a required review step.

78. Latency rose by 80 ms after a new auth plugin was enabled. What do you do?

Measure first: compare Kong's latency metrics (Kong's own time versus upstream time) before and after. An auth plugin that calls an external introspection or identity endpoint on every request is the usual cause. Enable the plugin's caching options, move to local JWT verification where security allows, and check network distance between the gateway and the IdP.

79. A security scan finds the Admin API reachable from the internet. What are your steps?

Block it immediately at the firewall or security group. Then review Admin API access logs and decK dumps for unexpected changes (new Consumers, plugins or Routes), rotate any credentials stored in Kong, rebind admin_listen to a private address, enable RBAC if you run Enterprise, and add a check in your infrastructure code that fails if ports 8001 or 8444 are public.

80. You need to move 200 Ingress resources from another ingress controller to Kong. How do you plan it?

Install KIC alongside the existing controller with its own IngressClass, so nothing changes until a resource opts in. Convert annotations from the old controller to Kong equivalents (plugins and konghq.com/* annotations), ideally moving to Gateway API HTTPRoutes at the same time. Migrate a few low-risk services first, compare logs and metrics, switch DNS per hostname, and keep the old controller until the last service moves.

How to prepare for a Kong interview

  • Run Kong locally and build one API end to end: Service, Route, key-auth, rate limit, Prometheus metrics.
  • Recreate the same setup as a decK file and as Kubernetes manifests with KIC, so you can talk about both.
  • Set up hybrid mode once with PKI certificates and break it on purpose (stop the CP, mismatch a version) to see the behaviour.
  • Practise the scenarios above aloud using symptom, check, fix and prevention.
  • Read the precedence list for plugin scopes and the rate-limiting policy table until you can explain them without notes.

For more practice, see our 75+ Kong interview questions and scenario-based Kong questions. DevOps Training Institute runs classroom batches in Kharadi, Pune and live online classes across India. The CKA course covers the Kubernetes networking and Ingress topics behind Kong Ingress Controller, and the DevOps course covers the CI/CD pipelines where decK runs.

FAQs

Is Kong Gateway free to use?

Kong publishes its gateway source code, and many features work without a licence. Enterprise features such as RBAC, Workspaces and several advanced plugins need a Kong Gateway Enterprise licence or Konnect. Check Kong's current packaging and pricing pages before you plan production use.

What is the difference between a Kong Service and a Route?

A Service describes the backend Kong sends traffic to: protocol, host, port and path. A Route describes which incoming requests go to that Service, by path, host, method or headers. One Service usually has several Routes.

Should I learn DB-less mode or hybrid mode first?

Start with DB-less mode, because it teaches the declarative file format with no database to manage. Then learn hybrid mode, which most production and Konnect deployments use and which interviewers ask about for experienced roles.

Can decK be used with DB-less mode?

Not to sync. Kong's docs state that decK cannot be used against a DB-less node, because its Admin API is read-only. You can still use decK file commands to build and check the declarative file, then load it through the /config endpoint or at startup.

Which Kong Gateway version should I mention in interviews?

Mention that you work on an LTS line and know the latest release. In November 2026 that meant 3.10 or 3.14 LTS, with 3.16 as the newest minor version. Interviewers care more that you understand the upgrade path than the exact number.

Is the Kong Ingress Controller required to run Kong on Kubernetes?

It is the standard way, because it turns Ingress, Gateway API and Kong custom resources into gateway configuration. You could run Kong pods managed only by decK, but you would lose the Kubernetes-native workflow most teams want.

What is the difference between Kong Gateway and Kong Konnect?

Kong Gateway is the proxy that handles traffic. Konnect is the hosted management platform that runs the control plane and adds Dev Portal, analytics, catalog and other services. Your data planes are Kong Gateway nodes in both cases.

How many Kong interview questions should I practise?

Be able to answer questions 1 to 25 without notes, write a decK file and a KongPlugin manifest from memory, and talk through at least six of the scenario questions in detail. That covers most fresher and mid-level rounds.

What's Your Reaction?

Like Like 0
Dislike Dislike 0
Love Love 0
Funny Funny 0
Angry Angry 0
Sad Sad 0
Wow Wow 0
Mridul I am a passionate technology enthusiast with a strong focus on DevOps, Cloud Computing, and Cybersecurity. Through my blogs at DevOps Training Institute, I aim to simplify complex concepts and share practical insights for learners and professionals. My goal is to empower readers with knowledge, hands-on tips, and industry best practices to stay ahead in the ever-evolving world of DevOps.